Among all the man-made sounds, speech is perhaps the most prolific and influential. It has attracted greatest attention, for it is perceived as a natural interface between man and machine—computers. Decades of research resulted in the emergence of Automatic Speech Recognition (ASR) as a leading technology serving as a man-machine interface. The central aim of speech recognition process is to match an input signal with a set of words according to some optimality criteria. And the first step in this process is parameterization—conversion of the input signal into parameters while preserving virtually all the speech signal information relating to the text message.

One of the most widely used tools for parametric representation of speech signal for speech recognition is the Mel Frequency Cepstrum Coefficients (MFCC). But it has a limitation: the signal is assumed to be stationary within the given time frame and so cannot analyze the non-stationary signal. To overcome this difficulty, several researchers have used different types of Amplitude Modulation (AM) and Frequency Modulation (FM) techniques for extracting features from speech signal. The authors, Vibha Tiwari and Jyoti Singhai, of the first paper, “AM-FM Features and Their Application to Noise Robust Speech Recognition: A Review”, have discussed various techniques using the AM-FM model for feature extraction of speech for speech recognition, besides investigating their efficiency levels. They have also studied the use of AM-FM modulation with Teager Energy Cepstral Coefficients (TECC). The results of the study indicate that the recognition efficiency goes up from 63% to 68%, when MFCC technique is used with AM demodulation. Recognition of female voice is found to be better than male voice. The efficiency goes up to 91.82%, if TECC is used.

Frequency selective fading occurs if the channel bandwidth of a communication system is larger than the coherent bandwidth of the channel. To reduce the intersymbol interference, multicarrier modulation method is resorted to using Orthogonal Frequency Division Multiplexing (OFDM). However OFDM is again prone to different interference signals. To obviate this, the authors, Manoj V J and Elizabeth Elias of the next paper, “Design of Low Complexity Cosine Modulated Filter Bank Transmultiplexer for Unknown Channels”, have designed a Low Complexity Cosine Modulated Filter Bank Transmultiplexer (LC CMFB TMUX). They have also synthesized the coefficients of the transmitting and receiving low-pass filters into a canonic signed digit format, using genetic algorithm. They have used modified crossover and mutation techniques to preserve the canonical property of the signed power of two representations. Simulation results indicate that LC CMFB TMUX designed by them has better signal to interference ratio and frequency responsors than those of LC CMFB TMUX obtained by rounding the coefficients of low-pass filters. To access a common resource simultaneously by two or more contenders, an arbiter is required. Asynchronous digital circuit designs are known to yield better results in terms of power efficiency, modularity, average-case performance, concurrency, reusability, noise immunity and metastability. The authors, Tamilarasi M, Shankar R, Jaweth Akther, Thirumala Rao, Manishekar C and Ramakrishnan N, of the next paper, “Design, Implementation and Performance Evaluation of Synchronous and Asynchronous Arbiters”, have designed a 4 ´ 1 asynchronous arbiter and evaluated its performance in comparison with a synchronous arbiter of the same specifications, as used for the design of the asynchronous arbiter. The measured performance parameters revealed that the asynchronous arbiter consumed less power, offered higher throughput, higher frequency of operation and lower latency vis-à-vis the synchronous 4 ´ 1 arbiter.

The authors, V K Dwivedi, S Tripathi, V S Tripathi, R Tripathi and S Tiwari, of the next paper, “Performance of a Variable Envelope Detector-Based Polar Transmitter for OFDM System”, have designed a new CMOS variable envelope detector-based polar transmitter for OFDM systems with an objective to minimize the average call drop of all users, that usually occurs due to fade in multipath environment, and evaluated its performance. Simulation results have shown that the algorithm proposed by them had improved the performance of the system significantly.

Conservation of storage space and fast data transmission are ensured using image compression techniques. Fractal image compression technique is known to provide very high compression ratio, fast decompression and resolution independence, but it is said to be very time consuming. The authors, Vijayshri Chaurasia and Ajay Somkuwar, of the next paper, “Fractal Image Compression with Approximation Error-Based Suitable Domain Search”, have designed a new approximation error-based domain search technique and evaluated its performance. The results indicated that the technique proposed reduced the time requirement of suitable domain search considerably.

The authors, K L Baishnab, Amlan Nag and F A Talukdar, of the last paper, “A High Speed Voltage Mode CMOS WTA Circuit for Image Processing”, have described a new design and simulation of a novel CMOS voltage mode Winner-Take-All (WTA) circuit that employs additional inhibitory and local excitatory feedback based on a common voltage computation, with a view to improve both speed and precision significantly. They have tested the circuit with five contending cells and the simulation results indicate a better dynamic range as well as resolution.

-- GRK Murty
Consulting Editor